CN114818549B - Method, system, equipment and medium for calculating fluid mechanics parameter of object - Google Patents
Method, system, equipment and medium for calculating fluid mechanics parameter of object Download PDFInfo
- Publication number
- CN114818549B CN114818549B CN202210752299.0A CN202210752299A CN114818549B CN 114818549 B CN114818549 B CN 114818549B CN 202210752299 A CN202210752299 A CN 202210752299A CN 114818549 B CN114818549 B CN 114818549B
- Authority
- CN
- China
- Prior art keywords
- field data
- flow field
- simulation
- equation
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000012530 fluid Substances 0.000 title claims abstract description 47
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000004088 simulation Methods 0.000 claims abstract description 125
- 238000013528 artificial neural network Methods 0.000 claims abstract description 79
- 230000007704 transition Effects 0.000 claims abstract description 26
- 238000012549 training Methods 0.000 claims description 32
- 230000006870 function Effects 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 description 14
- 238000013461 design Methods 0.000 description 13
- 238000004422 calculation algorithm Methods 0.000 description 10
- 238000013507 mapping Methods 0.000 description 6
- 238000012937 correction Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 238000003062 neural network model Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000012896 Statistical algorithm Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Abstract
The invention discloses a method for calculating fluid mechanics parameters of an object, which comprises the following steps: simulating under a plurality of simulation parameters by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, and simulating under corresponding simulation parameters by using a second RANS model equation and an NS control equation to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein a transition model is fused in the second RANS model equation; constructing a neural network, taking flow field data as input, and taking a corresponding vortex viscosity coefficient as output to train the neural network; acquiring a simulation object of fluid mechanics parameters to be calculated, and simulating the simulation object by using a first RANS model equation and an NS control equation to obtain flow field data; and inputting the flow field data into the trained neural network to obtain a corresponding vortex-viscosity coefficient, and calculating the fluid mechanics parameter of the simulation object by using the corresponding vortex-viscosity coefficient and the NS control equation. The invention also discloses a system, a computer device and a readable storage medium.
Description
Technical Field
The invention relates to the field of fluid mechanics, in particular to a method, a system, equipment and a storage medium for calculating fluid mechanics parameters of an object.
Background
With the development of efficient inference statistical algorithms and machine learning techniques, students in different fields have begun to try to use previously developed high confidence data (high precision simulation and experimental measurements, etc.) to assist in modeling. For Computational Fluid Dynamics (CFD), artificial Neural Network (ANN) -based algorithm-assisted turbulence model modeling has also become one of the current research hotspots.
The computational fluid mechanics means that a numerical simulation method is utilized to solve a basic control equation (Navier-Stokes, N-S equation) of fluid mechanics, and since the equation does not have an analytic solution due to the property of the equation, a certain model assumption needs to be given when the numerical equation is used for solving the equation. In any assumption, a certain approximation exists, so that the solution accuracy is affected, and a high-accuracy simulation result needs a large amount of computing resources to obtain. Therefore, data-driven assisted modeling may be a reasonable solution to the above-described contradiction.
In terms of engineering fluid mechanics simulation, the most widely applied method at present is to convert an N-S equation into a Reynolds average N-S equation (RANS) for solution, i.e., all physical quantities in the N-S equation are decomposed into time average quantity and pulse quantity and then are brought into the original equation to obtain the product. The RANS equation is very similar in form to the N-S equation, except that a term related to Reynolds stress is added, so a new equation needs to be added to make the RANS equation closed. Different RANS models can be obtained according to different numbers of the added equations. However, since the RANS equation itself mainly focuses on the change of the time-average values of different physical quantities, it has a defect in solving many problems, such as completely failing to capture a physical process, which is very important in fluid mechanics, transition (indicating the transition of the fluid state from laminar flow to turbulent flow). To solve this problem, many scholars implement the transition phenomenon capture by adding one to two new equations, namely a transition model.
The application of data-driven modeling based on neural network algorithm in computational fluid mechanics mainly has the following forms:
1. and evaluating and correcting the uncertainty of the turbulence model in the RANS method based on a neural network algorithm and high-confidence data. Because different turbulence models have certain deviation on the simulation of the Reynolds stress uncertainty in the RANS equation, the uncertainty can be quantified and corrected by a neural network algorithm.
2. And evaluating and correcting the to-be-determined coefficients and the sizes of different items in the turbulence model equation based on a neural network algorithm and high-confidence data. Similar to the first approach, the object of knowledge evaluation and correction focuses on the turbulence model equation itself.
However, the above-mentioned neural network algorithm has different defects in both methods commonly used in computational fluid dynamics simulation. That is, only prediction can be performed on the existing data set, and the prediction cannot be used for correcting the data set of the calculation example which does not appear in the model training process, that is, certain generalization is lacked. Furthermore, the transition phenomenon cannot be predicted.
Disclosure of Invention
In view of the above, in order to overcome at least one aspect of the above problems, an embodiment of the present invention provides a method for calculating a fluid mechanics parameter of an object, including the following steps:
simulating under a plurality of simulation parameters by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, and simulating under corresponding simulation parameters by using a second RANS model equation and the NS control equation to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein a transition model is fused in the second RANS model equation;
constructing a neural network, taking the flow field data as input, and taking a corresponding vortex viscosity coefficient as output to train the neural network;
acquiring a simulation object of fluid mechanics parameters to be calculated, and simulating the simulation object by using the first RANS model equation and the NS control equation to obtain flow field data;
and inputting the flow field data into a trained neural network to obtain a corresponding vortex-viscosity coefficient, and calculating the fluid mechanics parameter of the simulation object by using the corresponding vortex-viscosity coefficient and the NS control equation.
In some embodiments, the simulating with the first RANS model equation and the NS control equation under the plurality of simulation parameters obtains a plurality of flow field data, further comprising:
and simulating under different attack angles, different Mach numbers and/or different simulation objects by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, wherein each flow field data comprises a plurality of physical quantities.
In some embodiments, training the neural network with the flow field data as input and the corresponding vortex-viscosity coefficients as output further comprises:
selecting as input a number of physical quantities from the plurality of physical quantities embodying a difference between laminar flow and turbulent flow.
In some embodiments, the number of physical quantities is speed, pressure, temperature, and/or density.
In some embodiments, training the neural network with the flow field data as an input and the corresponding vortex-viscosity coefficients as an output further comprises:
and normalizing the selected physical quantities.
In some embodiments, the simulating with the first RANS model equation and the NS control equation under the plurality of simulation parameters obtains a plurality of flow field data, further comprising:
and simulating under a plurality of simulation parameters by using a first RANS model equation and an NS control equation under a first density grid to obtain a plurality of flow field data.
In some embodiments, the simulating with the second RANS model equation and the NS control equation under the corresponding simulation parameters obtains the vortex viscosity coefficient corresponding to each flow field data, further including:
and simulating under corresponding simulation parameters by using a second RANS model equation and the NS control equation under a second density grid to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein the second density grid is denser than the first density grid.
In some embodiments, training the neural network with the flow field data as input and the corresponding vortex-viscosity coefficients as output further comprises:
interpolating the corresponding vortex-viscosity coefficients obtained under the second density grid onto the first density grid and calculating an inverse hyperbolic sine function as output.
In some embodiments, constructing the neural network further comprises:
and constructing the neural network by utilizing an input layer, three hidden layers and an output layer.
In some embodiments, further comprising:
using the mean square error formula
Calculating a loss function, wherein y i To predict value, t i Are true values.
Based on the same inventive concept, according to another aspect of the present invention, an embodiment of the present invention further provides a fluid mechanics parameter calculation system of an object, comprising:
the simulation module is configured to perform simulation under a plurality of simulation parameters by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, and perform simulation under corresponding simulation parameters by using a second RANS model equation and the NS control equation to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein a transition model is fused in the second RANS model equation;
the training module is configured to construct a neural network, take the flow field data as input, and take the corresponding vortex viscosity coefficient as output to train the neural network;
the acquisition module is configured to acquire a simulation object of fluid mechanics parameters to be calculated, and simulate the simulation object by using the first RANS model equation and the NS control equation to obtain flow field data;
and the calculation module is configured to input the flow field data into a trained neural network to obtain a corresponding vortex-viscosity coefficient, and calculate the fluid mechanics parameter of the simulation object by using the corresponding vortex-viscosity coefficient and the NS control equation.
In some embodiments, the simulation module is further configured to:
and simulating under different attack angles, different Mach numbers and/or different simulation objects by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, wherein each flow field data comprises a plurality of physical quantities.
In some embodiments, the training module is further configured to:
selecting as input a number of physical quantities from the plurality of physical quantities embodying a difference between laminar flow and turbulent flow.
In some embodiments, the number of physical quantities is speed, pressure, temperature, and/or density.
In some embodiments, the training module is further configured to:
and carrying out normalization processing on the selected physical quantities.
In some embodiments, the simulation module is further configured to:
and simulating under a first density grid by utilizing a first RANS model equation and an NS control equation under a plurality of simulation parameters to obtain a plurality of flow field data.
In some embodiments, the simulation module is further configured to: and simulating under a second density grid by using a second RANS model equation and the NS control equation under corresponding simulation parameters to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein the second density grid is denser than the first density grid.
In some embodiments, the simulation module is further configured to:
interpolating the corresponding vortex-viscosity coefficients obtained under the second density grid onto the first density grid and calculating an inverse hyperbolic sine function as output.
In some embodiments, the training module is further configured to:
and constructing the neural network by utilizing an input layer, three hidden layers and an output layer.
In some embodiments, the training module is further configured to:
using the mean square error formula
Calculating a loss function, wherein y i Is a predicted value, t i Are true values.
Based on the same inventive concept, according to another aspect of the present invention, an embodiment of the present invention further provides a computer apparatus, including:
at least one processor; and
a memory storing a computer program operable on the processor, wherein the processor, when executing the program, performs the steps of any of the methods of calculating a fluid mechanical parameter of an object as described above.
Based on the same inventive concept, according to another aspect of the present invention, there is also provided a computer-readable storage medium storing a computer program which, when executed by a processor, performs the steps of the fluid mechanics parameter calculation method of any one of the objects as described above.
The invention has one of the following beneficial technical effects: according to the scheme provided by the invention, a new turbulence model is designed based on a neural network algorithm and flow field data considering the transition phenomenon, the model can replace the traditional RANS turbulence model and realize the capture of the transition phenomenon, the numerical simulation calculation amount is reduced, and the accuracy of a prediction result is improved. Turbulence simulation is widely applied to industries, such as design of wings or engine blades of airplanes, design of ship propellers, design of blades of wind driven generators and the like, and the turbulence model or the idea for optimizing the existing turbulence model provided by the invention can provide assistance for industrial design and simulation of different industries.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other embodiments can be obtained by using the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for calculating a fluid mechanical parameter of an object according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a fluid mechanics parameter calculation system for an object according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a computer device provided in an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a computer-readable storage medium according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the following embodiments of the present invention are described in further detail with reference to the accompanying drawings.
It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are used for distinguishing two entities with the same name but different names or different parameters, and it should be noted that "first" and "second" are merely for convenience of description and should not be construed as limitations of the embodiments of the present invention, and they are not described in any more detail in the following embodiments.
According to an aspect of the present invention, an embodiment of the present invention provides a fluid mechanics parameter calculation method for an object, as shown in fig. 1, which may include the steps of:
s1, simulating under a plurality of simulation parameters by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, and simulating under corresponding simulation parameters by using a second RANS model equation and the NS control equation to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein a transition model is fused in the second RANS model equation;
s2, constructing a neural network, taking the flow field data as input, and taking a corresponding vortex-viscosity coefficient as output to train the neural network;
s3, acquiring a simulation object of the fluid mechanics parameter to be calculated, and simulating the simulation object by using the first RANS model equation and the NS control equation to obtain flow field data;
and S4, inputting the flow field data into a trained neural network to obtain a corresponding vortex-viscosity coefficient, and calculating the fluid mechanics parameter of the simulation object by using the corresponding vortex-viscosity coefficient and the NS control equation.
According to the scheme provided by the invention, a new turbulence model is designed based on a neural network algorithm and flow field data considering the transition phenomenon, the model can replace the traditional RANS turbulence model and realize the capture of the transition phenomenon, the numerical simulation calculation amount is reduced, and the accuracy of a prediction result is improved. Turbulence simulation is widely applied to industries, such as design of wings or engine blades of airplanes, design of ship propellers, design of blades of wind driven generators and the like, and the turbulence model provided by the invention or the idea for optimizing the existing turbulence model can provide assistance for industrial design and simulation of different industries.
In some embodiments, simulating using the first RANS model equation and the NS control equation under the plurality of simulation parameters to obtain a plurality of flow field data further comprises:
and simulating under different attack angles, different Mach numbers and/or different simulation objects by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, wherein each flow field data comprises a plurality of physical quantities.
Specifically, a simulation result of a conventional RANS model (referred to as a RANS-Original) and a RANS model (referred to as a RANS-Transition) fused with a Transition model under different attack angles, different mach numbers and even different wings can be used.
In some embodiments, training the neural network with the flow field data as an input and the corresponding vortex-viscosity coefficients as an output further comprises:
selecting as input a number of physical quantities from the plurality of physical quantities embodying a difference between laminar flow and turbulent flow.
In some embodiments, the number of physical quantities is speed, pressure, temperature, and/or density.
Specifically, the transition phenomenon relates to the transition of the fluid state from laminar flow to turbulent flow, and in consideration of the physical phenomenon, a physical quantity capable of reflecting the difference between the laminar flow and the turbulent flow is selected as an input quantity of the neural network model, and meanwhile, partial tensor components are added to enhance the convergence of model training. For example, the physical quantities selected may be velocity, pressure, temperature, and/or density, among others.
In some embodiments, training the neural network with the flow field data as input and the corresponding vortex-viscosity coefficients as output further comprises:
and normalizing the selected physical quantities.
Specifically, a plurality of physical quantities I _ Y can be selected from RANS-Original simulation results of the training data set according to physical characteristics of simulation objects (such as wings and the like) train As input features for neural networks, used before training
And carrying out normalization processing by using a formula.
In some embodiments, simulating using the first RANS model equation and the NS control equation under the plurality of simulation parameters to obtain a plurality of flow field data further comprises:
and simulating under a plurality of simulation parameters by using a first RANS model equation and an NS control equation under a first density grid to obtain a plurality of flow field data.
In some embodiments, the simulating with the second RANS model equation and the NS control equation under the corresponding simulation parameters obtains the vortex viscosity coefficient corresponding to each flow field data, further including:
and simulating under corresponding simulation parameters by using a second RANS model equation and the NS control equation under a second density grid to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein the second density grid is denser than the first density grid.
In some embodiments, training the neural network with the flow field data as an input and the corresponding vortex-viscosity coefficients as an output further comprises:
interpolating the corresponding vortex-viscosity coefficients obtained under the second density grid onto the first density grid and calculating an inverse hyperbolic sine function as output.
Specifically, the inverse hyperbolic sinusoids arcsinh (μ) of the vortex viscosity coefficient can be extracted from the RANS-Transition data of the training set t true )| train The method has the advantages that the data of the dense grids are interpolated on the coarse grids to serve as output parameters, the stability of later iteration of the model is enhanced, namely the output value for training is a value with higher precision, so that the convergence and the stability in the simulation process are improved by using the mode of the dense grid interpolation, the physical phenomenon can be accurately simulated on the coarse grids, and the calculated amount is greatly saved.
In the selection of the output parameters, the reynolds stress is mainly determined by the viscosity coefficient mu of the turbulence t And finally, selecting an inverse hyperbolic sine function of the true value of the vortex viscosity coefficient as a final output parameter, namely arcsinh (mu), to avoid the deviation of local region prediction caused by overlarge numerical difference of the vortex viscosity coefficients in different regions (such as the surface of an object and the outer boundary of a simulation region) t true )。
Finally, a mapping based on the training set data is constructed, i.e.
And then deriving the weight and the offset of each neuron in the neural network under the mapping, embedding the obtained weight and the offset into an open source CFD code, namely predicting the vortex viscosity coefficient distribution of a solving problem by using a neural network model, and finally realizing the optimization of the simulation result of the RANS-origin model. And S3, acquiring a simulation object of the fluid mechanics parameter to be calculated, simulating the simulation object by using the first RANS model equation and the NS control equation to obtain flow field data, then S4, inputting the flow field data into a trained neural network to obtain a corresponding vortex viscosity coefficient, and calculating the fluid mechanics parameter of the simulation object by using the corresponding vortex viscosity coefficient and the NS control equation.
In some embodiments, constructing the neural network further comprises:
and constructing the neural network by utilizing an input layer, three hidden layers and an output layer.
In some embodiments, further comprising:
using the mean square error formula
Calculating a loss function, wherein y i To predict value, t i Are true values.
Specifically, keras can be used as a neural network framework building tool, is an open source artificial neural network library written by Python, and can be used as a high-level application program interface of Tensorflow, theano and the like. Currently, many open source CFD codes (for example, openFoam, etc.) have fused transition models, so that a certain open source CFD code can be selected as the second RANS model.
The hyper-parameter setting of the neural network framework is as follows:
activation function: the light of the Leaky ReLu,
wherein a is i Is a fixed parameter in the interval (1, + ∞).
An optimizer: the number of the Adam,
adam is a first-order optimization algorithm that can replace the traditional stochastic gradient descent process, and can iteratively update neural network weights based on training data.
Loss function: the mean square error of the measured values of the parameters,
and calculating the Euclidean distance between the predicted value and the real value, wherein the closer the predicted value and the real value are, the smaller the mean square error of the predicted value and the real value is.
And verifying the accuracy of the vortex-viscous coefficient predicted by the neural network model based on the flow field result after code convergence, and further optimizing the neural network parameters until the result is accurate if the deviation exists.
The scheme provided by the invention designs a neural network framework, establishes mapping between physical parameters of the traditional fluid mechanics turbulence model and vortex viscosity coefficients of the turbulence model considering transition phenomenon, and further realizes correction of a prediction result of the traditional turbulence model. In the process of establishing the neural network framework, the selection of input physical quantities and the form of output parameters are screened and improved by considering the physical characteristics of the simulated problem. And the convergence and stability in the simulation process are improved by using a close grid interpolation mode, so that the physical phenomenon can be accurately simulated on a thicker grid, and the calculated amount is greatly saved.
Based on the same inventive concept, according to another aspect of the present invention, an embodiment of the present invention further provides a fluid mechanics parameter calculation system 400 of an object, as shown in fig. 2, comprising:
the simulation module 401 is configured to perform simulation under multiple simulation parameters by using a first RANS model equation and an NS control equation to obtain multiple flow field data, and perform simulation under corresponding simulation parameters by using a second RANS model equation and the NS control equation to obtain a vortex viscosity coefficient corresponding to each flow field data, where a transition model is fused in the second RANS model equation;
a training module 402 configured to construct a neural network and train the neural network with the flow field data as input and a corresponding vortex-viscosity coefficient as output;
an obtaining module 403, configured to obtain a simulation object of the fluid mechanics parameter to be calculated, and simulate the simulation object by using the first RANS model equation and the NS control equation to obtain flow field data;
a calculating module 404 configured to input the flow field data into a trained neural network to obtain a corresponding vortex-viscosity coefficient and calculate a fluid mechanics parameter of the simulated object by using the corresponding vortex-viscosity coefficient and the NS control equation.
In some embodiments, the simulation module 401 is further configured to:
and simulating under different attack angles, different Mach numbers and/or different simulation objects by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, wherein each flow field data comprises a plurality of physical quantities.
In some embodiments, training module 402 is further configured to:
selecting as input a number of physical quantities from the plurality of physical quantities embodying a difference between laminar flow and turbulent flow.
In some embodiments, the number of physical quantities is speed, pressure, temperature, and/or density.
In some embodiments, the training module 402 is further configured to:
and carrying out normalization processing on the selected physical quantities.
In some embodiments, the simulation module 401 is further configured to:
and simulating under a plurality of simulation parameters by using a first RANS model equation and an NS control equation under a first density grid to obtain a plurality of flow field data.
In some embodiments, the simulation module 401 is further configured to: and simulating under a second density grid by using a second RANS model equation and the NS control equation under corresponding simulation parameters to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein the second density grid is denser than the first density grid.
In some embodiments, the simulation module 401 is further configured to:
interpolating the corresponding vortex-viscosity coefficients obtained under the second density grid onto the first density grid and calculating an inverse hyperbolic sine function as output.
In some embodiments, the training module 402 is further configured to:
and constructing the neural network by utilizing an input layer, three hidden layers and an output layer.
In some embodiments, training module 402 is further configured to:
using the mean square error formula
Calculating a loss function, wherein y i To predict value, t i Are true values.
The scheme provided by the invention designs a neural network framework, establishes mapping between physical parameters of the traditional fluid mechanics turbulence model and vortex viscosity coefficients of the turbulence model considering transition phenomenon, and further realizes correction of a prediction result of the traditional turbulence model. In the process of establishing the neural network framework, the selection of input physical quantities and the form of output parameters are screened and improved by considering the physical characteristics of the simulated problem. And the convergence and stability in the simulation process are improved by using a close grid interpolation mode, so that the physical phenomenon can be accurately simulated on a thicker grid, and the calculated amount is greatly saved.
Based on the same inventive concept, according to another aspect of the present invention, as shown in fig. 3, an embodiment of the present invention further provides a computer apparatus 501, comprising:
at least one processor 520; and
a memory 510, the memory 510 storing a computer program 511 executable on the processor, the processor 520 executing the program to perform the steps of:
s1, simulating under a plurality of simulation parameters by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, and simulating under corresponding simulation parameters by using a second RANS model equation and the NS control equation to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein a transition model is fused in the second RANS model equation;
s2, constructing a neural network, taking the flow field data as input, and taking a corresponding vortex-viscosity coefficient as output to train the neural network;
s3, acquiring a simulation object of the fluid mechanics parameter to be calculated, and simulating the simulation object by using the first RANS model equation and the NS control equation to obtain flow field data;
and S4, inputting the flow field data into a trained neural network to obtain a corresponding vortex-viscosity coefficient, and calculating the fluid mechanics parameter of the simulation object by using the corresponding vortex-viscosity coefficient and the NS control equation.
In some embodiments, the simulating with the first RANS model equation and the NS control equation under the plurality of simulation parameters obtains a plurality of flow field data, further comprising:
and simulating under different attack angles, different Mach numbers and/or different simulation objects by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, wherein each flow field data comprises a plurality of physical quantities.
In some embodiments, training the neural network with the flow field data as an input and the corresponding vortex-viscosity coefficients as an output further comprises:
selecting as input a number of physical quantities from the plurality of physical quantities embodying a difference between laminar flow and turbulent flow.
In some embodiments, the number of physical quantities is speed, pressure, temperature, and/or density.
In some embodiments, training the neural network with the flow field data as input and the corresponding vortex-viscosity coefficients as output further comprises:
and normalizing the selected physical quantities.
In some embodiments, the simulating with the first RANS model equation and the NS control equation under the plurality of simulation parameters obtains a plurality of flow field data, further comprising:
and simulating under a first density grid by utilizing a first RANS model equation and an NS control equation under a plurality of simulation parameters to obtain a plurality of flow field data.
In some embodiments, the simulating with the second RANS model equation and the NS control equation under the corresponding simulation parameters obtains the vortex viscosity coefficient corresponding to each flow field data, further including:
and simulating under a second density grid by using a second RANS model equation and the NS control equation under corresponding simulation parameters to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein the second density grid is denser than the first density grid.
In some embodiments, training the neural network with the flow field data as an input and the corresponding vortex-viscosity coefficients as an output further comprises:
interpolating the corresponding vortex-viscosity coefficients obtained under the second density grid onto the first density grid and calculating an inverse hyperbolic sine function as output.
In some embodiments, constructing the neural network further comprises:
and constructing the neural network by utilizing an input layer, three hidden layers and an output layer.
In some embodiments, further comprising:
using the mean square error formula
Calculating a loss function, wherein y i Is a predicted value, t i Are true values.
The scheme provided by the invention designs a neural network framework, establishes mapping between physical parameters of the traditional fluid mechanics turbulence model and vortex viscosity coefficients of the turbulence model considering transition phenomenon, and further realizes correction of a prediction result of the traditional turbulence model. In the process of establishing the neural network framework, the selection of input physical quantities and the form of output parameters are screened and improved by considering the physical characteristics of the simulated problem. And the convergence and stability in the simulation process are improved by using a close grid interpolation mode, so that the physical phenomenon can be accurately simulated on a thicker grid, and the calculated amount is greatly saved.
Based on the same inventive concept, according to another aspect of the present invention, as shown in fig. 4, an embodiment of the present invention further provides a computer-readable storage medium 601, the computer-readable storage medium 601 stores a computer program 610, and the computer program 610 performs the following steps when executed by a processor:
s1, simulating under a plurality of simulation parameters by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, and simulating under corresponding simulation parameters by using a second RANS model equation and the NS control equation to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein a transition model is fused in the second RANS model equation;
s2, constructing a neural network, taking the flow field data as input, and taking a corresponding vortex viscosity coefficient as output to train the neural network;
s3, acquiring a simulation object of the fluid mechanics parameter to be calculated, and simulating the simulation object by using the first RANS model equation and the NS control equation to obtain flow field data;
and S4, inputting the flow field data into a trained neural network to obtain a corresponding vortex-viscosity coefficient, and calculating the fluid mechanics parameter of the simulation object by using the corresponding vortex-viscosity coefficient and the NS control equation.
In some embodiments, simulating using the first RANS model equation and the NS control equation under the plurality of simulation parameters to obtain a plurality of flow field data further comprises:
and simulating under different attack angles, different Mach numbers and/or different simulation objects by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, wherein each flow field data comprises a plurality of physical quantities.
In some embodiments, training the neural network with the flow field data as an input and the corresponding vortex-viscosity coefficients as an output further comprises:
selecting as input a number of physical quantities from the plurality of physical quantities embodying a difference between laminar flow and turbulent flow.
In some embodiments, the number of physical quantities is speed, pressure, temperature, and/or density.
In some embodiments, training the neural network with the flow field data as an input and the corresponding vortex-viscosity coefficients as an output further comprises:
and carrying out normalization processing on the selected physical quantities.
In some embodiments, the simulating with the first RANS model equation and the NS control equation under the plurality of simulation parameters obtains a plurality of flow field data, further comprising:
and simulating under a first density grid by utilizing a first RANS model equation and an NS control equation under a plurality of simulation parameters to obtain a plurality of flow field data.
In some embodiments, the simulating with the second RANS model equation and the NS control equation under the corresponding simulation parameters obtains the vortex viscosity coefficient corresponding to each flow field data, further including:
and simulating under corresponding simulation parameters by using a second RANS model equation and the NS control equation under a second density grid to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein the second density grid is denser than the first density grid.
In some embodiments, training the neural network with the flow field data as input and the corresponding vortex-viscosity coefficients as output further comprises:
interpolating the corresponding vortex-viscosity coefficients obtained under the second density grid onto the first density grid and calculating an inverse hyperbolic sine function as output.
In some embodiments, constructing the neural network further comprises:
and constructing the neural network by utilizing an input layer, three hidden layers and an output layer.
In some embodiments, further comprising:
using the mean square error formula
Calculating a loss function, wherein y i To predict value, t i Are true values.
The scheme provided by the invention designs a neural network framework, establishes mapping between physical parameters of the traditional fluid mechanics turbulence model and vortex viscosity coefficients of the turbulence model considering transition phenomenon, and further realizes correction of a prediction result of the traditional turbulence model. In the process of establishing the neural network framework, the selection of input physical quantities and the form of output parameters are screened and improved by considering the physical characteristics of the simulated problem. And the convergence and stability in the simulation process are improved by using a close grid interpolation mode, so that the physical phenomenon can be accurately simulated on a thicker grid, and the calculated amount is greatly saved.
Finally, it should be noted that, as will be understood by those skilled in the art, all or part of the processes of the methods of the above embodiments may be implemented by a computer program, which may be stored in a computer-readable storage medium, and when executed, may include the processes of the embodiments of the methods described above.
Further, it should be appreciated that the computer-readable storage media (e.g., memory) herein can be either volatile memory or nonvolatile memory, or can include both volatile and nonvolatile memory.
Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the disclosure herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as software or hardware depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the disclosed embodiments of the present invention.
The foregoing is an exemplary embodiment of the present disclosure, but it should be noted that various changes and modifications could be made herein without departing from the scope of the present disclosure as defined by the appended claims. The functions, steps and/or actions of the method claims in accordance with the disclosed embodiments described herein need not be performed in any particular order. Furthermore, although elements of the embodiments of the invention may be described or claimed in the singular, the plural is contemplated unless limitation to the singular is explicitly stated.
It should be understood that, as used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly supports the exception. It should also be understood that "and/or" as used herein is meant to include any and all possible combinations of one or more of the associated listed items.
The numbers of the embodiments disclosed in the embodiments of the present invention are merely for description, and do not represent the merits of the embodiments.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
Those of ordinary skill in the art will understand that: the discussion of any embodiment above is meant only to be exemplary, and is not intended to intimate that the scope of the disclosure, including the claims, is limited to these examples; within the idea of an embodiment of the invention, also technical features in the above embodiment or in different embodiments may be combined and there are many other variations of the different aspects of the embodiments of the invention as described above, which are not provided in detail for the sake of brevity. Therefore, any omissions, modifications, substitutions, improvements, and the like that may be made without departing from the spirit and principles of the embodiments of the present invention are intended to be included within the scope of the embodiments of the present invention.
Claims (15)
1. A method of calculating a fluid-mechanical parameter of an object, comprising the steps of:
simulating under a plurality of simulation parameters by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, and simulating under corresponding simulation parameters by using a second RANS model equation and the NS control equation to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein a transition model is fused in the second RANS model equation;
constructing a neural network, taking the flow field data as input, and taking a corresponding vortex viscosity coefficient as output to train the neural network;
acquiring a simulation object of fluid mechanics parameters to be calculated, and simulating the simulation object by using the first RANS model equation and the NS control equation to obtain flow field data;
and inputting the flow field data into a trained neural network to obtain a corresponding vortex-viscosity coefficient, and calculating the fluid mechanics parameter of the simulation object by using the corresponding vortex-viscosity coefficient and the NS control equation.
2. The method of claim 1, wherein simulating using the first RANS model equation and the NS control equation under the plurality of simulation parameters results in a plurality of flow field data, further comprising:
and simulating under different attack angles, different Mach numbers and/or different simulation objects by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, wherein each flow field data comprises a plurality of physical quantities.
3. The method of claim 2, wherein the neural network is trained using the flow field data as an input and a corresponding vortex-viscosity coefficient as an output, further comprising:
selecting as input a number of physical quantities from the plurality of physical quantities embodying a difference between laminar flow and turbulent flow.
4. A method according to claim 3, characterized in that said several physical quantities are speed, pressure, temperature and/or density.
5. The method of claim 3, wherein the neural network is trained using the flow field data as an input and the corresponding vortex-viscosity coefficients as an output, further comprising:
and carrying out normalization processing on the selected physical quantities.
6. The method of claim 1, wherein simulating using the first RANS model equation and the NS control equation under the plurality of simulation parameters results in a plurality of flow field data, further comprising:
and simulating under a first density grid by utilizing a first RANS model equation and an NS control equation under a plurality of simulation parameters to obtain a plurality of flow field data.
7. The method of claim 6, wherein the vortex-viscosity coefficient corresponding to each flow field data is obtained by performing a simulation using a second RANS model equation and the NS control equation under corresponding simulation parameters, further comprising:
and simulating under corresponding simulation parameters by using a second RANS model equation and the NS control equation under a second density grid to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein the second density grid is denser than the first density grid.
8. The method of claim 7, wherein the neural network is trained using the flow field data as an input and a corresponding vortex-viscosity coefficient as an output, further comprising:
interpolating the corresponding vortex-viscosity coefficients obtained under the second density grid onto the first density grid and calculating an inverse hyperbolic sine function as output.
9. The method of claim 1, wherein constructing a neural network further comprises:
and constructing the neural network by utilizing an input layer, three hidden layers and an output layer.
10. The method of claim 9, further comprising:
using the mean square error formula
Calculating a loss function, wherein y i To predict value, t i Is the true value;
wherein n is the number of predicted values or actual values, and i is the number of predicted values or actual values.
11. A fluid mechanics parameter calculation system for an object, comprising:
the simulation module is configured to perform simulation under a plurality of simulation parameters by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, and perform simulation under corresponding simulation parameters by using a second RANS model equation and the NS control equation to obtain a vortex viscosity coefficient corresponding to each flow field data, wherein a transition model is fused in the second RANS model equation;
the training module is configured to construct a neural network, take the flow field data as input, and take the corresponding vortex-viscosity coefficient as output to train the neural network;
the acquisition module is configured to acquire a simulation object of fluid mechanics parameters to be calculated, and simulate the simulation object by using the first RANS model equation and the NS control equation to obtain flow field data;
and the calculation module is configured to input the flow field data into a trained neural network to obtain a corresponding vortex-viscosity coefficient and calculate the fluid mechanics parameter of the simulation object by using the corresponding vortex-viscosity coefficient and the NS control equation.
12. The system of claim 11, wherein the simulation module is further configured to:
and simulating under different attack angles, different Mach numbers and/or different simulation objects by using a first RANS model equation and an NS control equation to obtain a plurality of flow field data, wherein each flow field data comprises a plurality of physical quantities.
13. The system of claim 12, wherein the training module is further configured to:
selecting as input a number of physical quantities from the plurality of physical quantities embodying a difference between laminar flow and turbulent flow.
14. A computer device, comprising:
at least one processor; and
memory storing a computer program operable on the processor, wherein the processor executes the program to perform the steps of the method according to any of claims 1-10.
15. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1-10.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210752299.0A CN114818549B (en) | 2022-06-29 | 2022-06-29 | Method, system, equipment and medium for calculating fluid mechanics parameter of object |
| PCT/CN2022/141089 WO2024001096A1 (en) | 2022-06-29 | 2022-12-22 | Method and system for calculating fluid mechanics parameters of object, and device and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210752299.0A CN114818549B (en) | 2022-06-29 | 2022-06-29 | Method, system, equipment and medium for calculating fluid mechanics parameter of object |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114818549A CN114818549A (en) | 2022-07-29 |
| CN114818549B true CN114818549B (en) | 2022-11-29 |
Family
ID=82522900
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210752299.0A Active CN114818549B (en) | 2022-06-29 | 2022-06-29 | Method, system, equipment and medium for calculating fluid mechanics parameter of object |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN114818549B (en) |
| WO (1) | WO2024001096A1 (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114818549B (en) * | 2022-06-29 | 2022-11-29 | 苏州浪潮智能科技有限公司 | Method, system, equipment and medium for calculating fluid mechanics parameter of object |
| CN116050242A (en) * | 2022-11-15 | 2023-05-02 | 中国空气动力研究与发展中心计算空气动力研究所 | Transition prediction method, device, equipment and medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112084727A (en) * | 2020-10-26 | 2020-12-15 | 中国人民解放军国防科技大学 | Transition prediction method based on neural network |
| CN113420507A (en) * | 2021-07-08 | 2021-09-21 | 刘维烁 | Fluid flow simulation method, device and medium |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7251592B1 (en) * | 2003-08-18 | 2007-07-31 | United Technologies Corporation | Boundary layer transition model |
| CN109033525B (en) * | 2018-06-27 | 2022-08-30 | 浙江大学 | Hypersonic transition prediction method based on simplified three-equation transition model |
| CN114818549B (en) * | 2022-06-29 | 2022-11-29 | 苏州浪潮智能科技有限公司 | Method, system, equipment and medium for calculating fluid mechanics parameter of object |
-
2022
- 2022-06-29 CN CN202210752299.0A patent/CN114818549B/en active Active
- 2022-12-22 WO PCT/CN2022/141089 patent/WO2024001096A1/en unknown
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112084727A (en) * | 2020-10-26 | 2020-12-15 | 中国人民解放军国防科技大学 | Transition prediction method based on neural network |
| CN113420507A (en) * | 2021-07-08 | 2021-09-21 | 刘维烁 | Fluid flow simulation method, device and medium |
Non-Patent Citations (2)
| Title |
|---|
| 基于方、矩形环管直接数值模拟结果的各向异性RANS湍流模型构建;郭永涛等;《复旦学报(自然科学版)》;20190228;第1-13页 * |
| 机器学习在湍流模型构建中的应用进展;张伟伟等;《空气动力学学报》;20190630;第444-454页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114818549A (en) | 2022-07-29 |
| WO2024001096A1 (en) | 2024-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114818549B (en) | Method, system, equipment and medium for calculating fluid mechanics parameter of object | |
| Cai et al. | Application of Bayesian networks in reliability evaluation | |
| Adeli et al. | Dynamic fuzzy wavelet neural network model for structural system identification | |
| CN111127246A (en) | Intelligent prediction method for transmission line engineering cost | |
| CN114492211B (en) | Residual oil distribution prediction method based on autoregressive network model | |
| Usman et al. | Machine learning computational fluid dynamics | |
| WO2021163309A1 (en) | Frequency-compensted neural networks for fluid dynamics design problems | |
| US20230359788A1 (en) | Simulating physical environments using graph neural networks | |
| CN115935834A (en) | History fitting method based on deep autoregressive network and continuous learning strategy | |
| CN111415010A (en) | Bayesian neural network-based wind turbine generator parameter identification method | |
| JP6973377B2 (en) | Analyst, analysis method, and program | |
| CN115879335A (en) | Fluid multi-physical-field parameter prediction method based on graph-generated neural network | |
| Blaud et al. | From multi‐physics models to neural network for predictive control synthesis | |
| Bao et al. | Study of data-driven mesh-model optimization in system thermal-hydraulic simulation | |
| Zargar et al. | Deep learning in early-stage structural performance prediction: Assessing morphological parameters for buildings | |
| CN114692529B (en) | CFD high-dimensional response uncertainty quantification method and device, and computer equipment | |
| CN116415482A (en) | Wing flow field analysis method based on graph neural network MeshGraphNet | |
| Wu et al. | Model Healing: Toward a framework for building designs to achieve code compliance | |
| US20200175121A1 (en) | System and method for predicting analytical abnormality in computational fluid dynamics analysis | |
| CN113887026A (en) | Oil pipeline hydraulic prediction method based on ensemble learning | |
| Corsini et al. | Machine learnt synthetic turbulence for LES inflow conditions | |
| Elloumi et al. | An overview on modelling of complex interconnected nonlinear systems | |
| Behnejad et al. | Deep learning in early-stage structural performance prediction: assessing morphological parameters for buildings | |
| US11922314B1 (en) | Systems and methods for building dynamic reduced order physical models | |
| CN111582632B (en) | Multi-factor safety stage prediction method for whole process of underground large space construction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |